Decision Support
The value of virtual pooling in dual sales channel supply chains

https://doi.org/10.1016/j.ejor.2007.09.034Get rights and content

Abstract

Recently the most significant growth in online retailing has been attributed to traditional offline retailers extending their brands online. Unfortunately, there is little research addressing the value of better information in retail/e-tail organizations. To fill this gap, this paper examines how investing in the continuous monitoring of online demands and inventory positions can provide economic benefit for companies that handle both in-store and online sales. Specifically, we develop and evaluate two dynamic assignment policies that incorporate real time information to specify which of a firm’s e-fulfillment locations will handle each of its Internet sales. Computational results indicate that investing in dynamic assignment capability can reduce system cost (holding, backorder, and transportation) by as much as 8.2% over the optimal static policy. The percentage of sales occurring online plays a critical role in determining the magnitude of the benefit.

Introduction

In today’s highly competitive business environment timely information presents a potential source of advantage for many companies. This is especially true in retail/e-tail organizations where such information serves as the glue for integrating online and offline operations (Laudon and Traver, 2003). According to an annual study by Forrester Research (2003), 63% of the 130 multi-channel retailers surveyed have upgraded to inventory management systems and 40% offer in-store inventory availabilities through their websites. Unfortunately, there is little research examining the value of such real time information in organizations that handle both in-store and online demand. This paper contributes to the literature by filling this gap and addressing the following two questions: (1) How can retailer/e-tailers leverage their central/real time supply chain information to provide economic benefit? (2) What is the magnitude of this benefit for a particular system configuration?

Anecdotal evidence suggests that even though numerous retail/e-tail organizations have added inventory management systems many are not leveraging the information that these systems provide. For example, traditional grocers like Albertsons and Safeway have recently begun selling groceries online. However, when customers shop at Safeway.com and enter their zip code they are instantly assigned to a store from which their order will be processed (Soars, 2003). Similarly, large appliance orders placed online at Lowe’s are assigned to the customer’s “local Lowe’s store” (www.lowes.com) while numerous other retailers allocate Internet sales to fulfillment centers based solely on their proximity to the online customer. Such static assignment policies pre-specify which location is responsible for handling online sales from each region. Unfortunately static assignments do not account for current inventory positions at the e-fulfillment locations and thereby increase the need to backorder or expedite online sales.

This paper examines how real time information on inventory positions and online demands can be leveraged for economic advantage in dual channel supply chains by dynamically specifying which e-fulfillment location will handle each Internet sale. Under dynamic assignment, online fulfillment responsibilities are determined in real time for each incoming online sale. Two such dynamic assignment policies are developed. Under each, online inventories are “virtually pooled” regardless of where they are physically positioned since online demand in one region can be met from inventory at a site in any other region. However, inventories dedicated to in-store sales are not virtually pooled since each in-store sale must be met (or backordered) via available inventory at the site where that sale originated.

To benchmark the “virtual pooling” benefits that can be obtained through dynamic assignment, the branch-and-bound algorithm developed in Bretthauer et al. (2006) is used to identify the optimal set of static assignments. Although this provides a baseline for comparison, Bretthauer et al. (2006) do not address virtual pooling or dynamic assignment in any way. The magnitude of the return that can be achieved by investing in the continuous monitoring of retail inventory positions is then evaluated by comparing system performance under the dynamic rules to: (1) that obtained under the optimal static policy; and (2) a lower bound on system cost.

To gain insight into static and dynamic assignment problems, we begin with a brief overview of relevant literature in the area of stochastic inventory control. Next, the static and dynamic assignment rules are developed. Computational results from a set of test problems are then analyzed to assess the value of real time information provided by inventory management systems. In the final section, we present some concluding remarks and directions for future research.

Section snippets

Literature review

In contrast to the substantial quantity of inventory control literature, research on multi-sales channel inventory systems is relatively sparse. An overview of analytical research models for dual sales channel supply chains is provided by Swaminathan and Tayur (2003). This survey illustrates some of the models that are currently being applied in dual sales channel settings to analyze procurement, information sharing, pricing, distribution, and customization decisions. The ideas in the paper

Assignment rules

In this section, an optimal static assignment policy and two dynamic assignment policies are developed to specify which of a retailer/e-tailer’s sites will handle each online sale. Consistent with the online fulfillment literature, we assume that channel demands are normally distributed (Bendoly et al., 2007, Bretthauer et al., 2006, Alptekinoğlu and Tang, 2005).

Simulation methodology for computing total cost

To compute the expected system-wide cost for a decentralized retailer/e-tailer operating under a dynamic or static assignment policy the general 2-echelon system in Fig. 1 is modeled in a computer simulation. Each period we assume that the sequence of events described in Section 3.1 takes place. Each online sale is allocated upon receipt according to the assignment policy in place. The dynamic assignment rules are evaluated by comparing the simulated costs under each rule to those obtained

Experimental study and test problems

To evaluate the proposed dynamic assignment rules and illustrate the degree to which dynamically specifying e-fulfillment responsibilities can improve system-wide cost versus an optimal static policy, a set of test problems are now developed based on a 14 location retailer/e-tailer operating in San Francisco, CA.

The problem setting was constructed by superimposing the firm’s 14 store locations on a discretized map of San Francisco (see Fig. 2). In each period, normally distributed stationary

Results and discussion

The expected system cost was obtained for each of the 180 factor combinations previously discussed, resulting in 900 runs. Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7, Fig. 8 illustrate that dynamically handling online sales can be used to leverage a company’s central/real time information and improve system performance.

Fig. 3 shows that for this problem set dynamic assignment reduces total holding, backorder, and transportation cost by as much as 8.2% over the optimal static policy. As the percent

Correlated demand and uncertain supplier lead time

A computational study was undertaken to determine if dynamic assignment rules provide good solutions even when demand is correlated and supplier lead times are uncertain. To assess the impact of demand correlation, the simulation study described in Section 5 was rerun under: (i) correlated in-store (online) demands between locations (regions) and (ii) correlated in-store and online demands within regions. Demand correlation was examined at five levels: positively correlated (i) and (ii),

Conclusions

In this paper, we have shown how retailer/e-tailers can leverage their central/real time supply chain information by dynamically specifying which e-fulfillment location will handle each Internet sale. Our results indicate that dynamic assignment (that is, “virtually pooling” online inventories) can substantially reduce total system cost over applying the optimal static policy. In our test problems the simple dynamic assignment policies reduced the total system cost of holding inventories,

References (30)

  • Bretthauer, K., Mahar, S., Venkataramanan, M.A., 2006. Solution methods for designing retail/e-tail distribution...
  • P. Chang et al.

    On the effect of centralization on expected costs in a multi-location newsboy problem

    The Journal of the Operational Research Society

    (1991)
  • M. Dada

    A two-echelon inventory system with priority shipments

    Management Science

    (1992)
  • B.M. De Koster

    Distribution strategies for online retailers

    IEEE Transactions on Engineering Management

    (2003)
  • G. Eppen

    Effects of centralization of expected costs in a multi-location newsboy problem

    Management Science

    (1979)
  • Cited by (68)

    • An inventory model with discounts for omnichannel retailers of slow moving items

      2022, European Journal of Operational Research
      Citation Excerpt :

      E-fulfillment and distribution strategies together with inventory integration are key aspects for cost reduction and high customer satisfaction for these retailers. Mahar et al. (2009) have shown the importance of assigning orders to fulfillment centers in a dynamic way. Fulfillment decisions become very complex when orders might contain multiple items, having different availability at different locations.

    View all citing articles on Scopus
    View full text